New sensor storage model

[kristopolous]

Lat/Lng matters for

• campaign provability
• insurance
• historical checking (eg, crime)

Voltage/current/temp matters for

• screen stability
• power model

6 DOF matters for

• collisions
• TBD driving habits

[kristopolous]

perhaps splitting up the sensors table to separate tables is appropriate

[kristopolous]

the tables could have distinctly different meta information. For example,

The 6dof could get the user-id of the current person driving the vehicle and so then we could easily break down responsibility of vehicle operation

The voltage/current/temp could store the software version for instance so we could see if one version is performing differently than another, etc.

The lat/lng could store campaign ids so we could easily figure out historicals from another source if the server had a glitch in storing them

[nthobe]

Proposed new sensor storage model

Store Location data in it's own table.

• location
  • id
  • lat (latitude)
  • lng (longitude)
  • accuracy (meters)
  • speed (kph)
  • heading (degrees relative to true north)
  • altitude (meters)
  • satellites (number of satellites used by gps)
  • hdop (accuracy multiplier)
  • source (gps, wifi)
  • time (seconds since epoch, UTC)

Sample and store every 5 seconds when using GPS. Every 5 minutes when using Wifi. Trigger on location table could update values in kv table for backwards compatibility.

Store motion sensor data in its own table along with a detected events table.

• sixdof
  • id
  • pitch
  • roll
  • yaw
  • accel_x
  • accel_y
  • accel_z
  • gyro_x
  • gyro_y
  • gyro_z
  • event_id (id of entry in events table)
  • time (seconds since epoch, UTC)
• events
  • id
  • type (crash, extreme braking, swerving, etc.)
  • confidence (percentage confidence in match)
  • time (seconds since epoch, UTC)

Sample data at 25Hz and save into db. Analyze data for events (accident, aggressive driving, etc) and save event information into the events table and set 'event_id' for samples before/during/after event. Regularly purge data older than our analysis buffer with null event_id.

Store state of screen in its own table.

• screen
  • id
  • backlight (brightness level)
  • fan (fan speed)
  • screen_sleep (boolean)
  • time (seconds since epoch, UTC)

Store state on SensorDaemon.py start and any time a value changes.

Store power/temp sensors in their own table.

• sensors
  • id
  • voltage (voltage reading)
  • current (current reading)
  • temp (temperature C)
  • t_read (thermistor reading)
  • t_res (thermistor resistance)

Sample data every second and store into db.

[nthobe]

New sensor storage model

Store Location data in it's own table.

• location
  • id
  • lat (latitude)
  • lng (longitude)
  • accuracy (meters)
  • speed (kph)
  • heading (degrees relative to true north)
  • source (gps, wifi)
  • created_at (datetime)

Sample and store every 5 seconds when using GPS. Every 5 minutes when using Wifi. The latest location is also stored in kv table as before.

Store motion sensor data in CSV file

• sixdof.csv
  • Time (Seconds from epoch UTC)
  • Accel_x (raw reading)
  • Accel_y (raw reading)
  • Accel_z (raw reading)
  • Gyro_x (raw reading)
  • Gyro_y (raw reading)
  • Gyro_z (raw reading)
  • Pitch (raw reading)
  • Roll (raw reading)
  • Yaw (raw reading)

Sample data at 25Hz and save into CSV file.

Store power/temp sensor data in CSV file

• powertemp.csv
  • Time (Seconds from epoch UTC)
  • Voltage (voltage reading)
  • Current (current reading)
  • Temp (temperature C)
  • Fan (fan pwm)
  • Light (backlight brightness pwm)
  • DPMS1 (On/Off for HDMI-1)
  • DPMS2 (On/Off for HDMI-2)

Sample data every 2 seconds and save into CSV file.

[nthobe]

Implemented and merged into master